MIT Develops Processors That Assemble Themselves

Photolithography, the current technology used in manufacturing processors, is starting to reach its limits and, as such, researchers are hard at work trying to find new procedures that can enable chip features to be shrunken even further. One of the technologies being explored is what MIT calls electron-beam lithography, a process that essentially uses electron beams instead of light to transfer mask patterns to layers of photoresist. The goal of the research is to eventually create chips made up of polymers and nano-posts that arrange themselves into useful patterns all on their own.

Photolitography builds chips layer by layer. A chip is covered in a layer of silicon, metal or some other material and is then coated with a photoresist, a light-sensitive material. Light is then projected on this photoresist, through a kind of “mask” (stencil), which projects a detailed pattern that hardens the photoresist when exposed.

The main issue with this technology is that today's chip features are actually smaller than the wavelength of the light used to make them, and the tricks developed by manufacturers to make that light produce smaller patterns can't work on smaller scales. In seeking to develop a solution for this limitation, the MIT researchers came up with the aforementioned electron-beam lithography.

The new procedure makes patterns of nano-posts on a silicon chip and deposits them with special polymers. This operation leads to certain hookups being formed between the posts and the polymer, which then arrange into useful patterns all on their own. These patterns can be very varied and quite useful in designing circuitry. For example, magnetic nanoscale patterns could be imprinted onto hard disks.

Of course, as with all technologies still in development, there is still a long way to go before self-arranging chips become practical. One issue is the fact that the electron-beam lithography is significantly more expensive than photolithography, because the scanning is much slower and more expensive. The researchers will focus on finding patterns that will produce functioning circuits in prototype processors.